Mixture of Experts: Advanced Architecture, Training, and Scaling
Chapter 1: Foundations of Sparse Expert Models
Conditional Computation Principles
The Sparse MoE Approach
Contrasting Dense vs. Sparse Activation
Mathematical Formulation of Basic MoE Layers
Chapter 2: Advanced MoE Architectures
Designing Effective Gating Networks
Hierarchical MoE Structures
Router Architectures: Linear, Non-Linear, Attention-Based
Expert Capacity and Sizing Considerations
Stabilization Techniques for Routers
Hands-on Practical: Implementing Custom Gating Mechanisms
Chapter 3: Training Dynamics and Optimization
The Load Balancing Problem in MoE
Auxiliary Loss Functions for Load Balancing
Router Optimization Strategies
Handling Dropped Tokens
Expert Specialization Collapse and Prevention
Impact of Optimizer Choice and Hyperparameters
Hands-on Practical: Implementing and Tuning Load Balancing Losses
Chapter 4: Scaling MoE Models: Distributed Training
Challenges in Distributed MoE Training
Expert Parallelism: Distributing Experts Across Devices
Integrating Expert Parallelism with Data Parallelism
All-to-All Communication Patterns
Pipeline Parallelism for MoE Models
Communication Optimization Techniques (e.g., Overlapping)
Frameworks and Libraries for Distributed MoE (e.g., DeepSpeed, Tutel)
Practice: Configuring Distributed MoE Training
Chapter 5: Inference Optimization and Deployment
Inference Challenges with Sparse Models
Batching Strategies for MoE Inference
Model Compression Techniques for MoE
Hardware Acceleration Considerations
Router Caching and Optimization
Deployment Patterns for Large Sparse Models
Hands-on Practical: Profiling MoE Inference
Conditional Computation Principles
Was this section helpful?
- Adaptive Mixture of Local Experts, Robert A. Jacobs, Michael I. Jordan, Steven J. Nowlan, and Geoffrey E. Hinton, 1991 Neural Computation, Vol. 3 (MIT Press) DOI: 10.1162/neco.1991.3.1.79 - Introduces the fundamental concept of a Mixture of Experts architecture where different 'experts' specialize in different regions of the input space, managed by a gating network.
- Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer, Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean, 2017 arXiv preprint arXiv:1701.06538 DOI: 10.48550/arXiv.1701.06538 - Presents the sparsely-gated Mixture-of-Experts layer, a key innovation for scaling neural networks to vast parameter counts while maintaining constant computational cost per example, particularly in the context of Transformers.
- GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding, Dmitry Lepikhin, Hieu Pham, Anselm Levskaya, Jonathan Shlens, Artem Grygoryev, Xingguang Chen, Yanqi Zhou, Yuanzhong Xu, Vikram Muralidharan, George Tucker, Anirudh Gowthaman, Chandraekhar Sowrirajan, David So, Jeffrey Dean, 2021 International Conference on Learning Representations (ICLR) 2021 DOI: 10.48550/arXiv.2006.16668 - Details a system for automatically sharding and training giant conditional computation models, including Sparse MoEs, across thousands of accelerators, enabling models with trillions of parameters.